Sistema de diagnóstico da esquistossomose a partir de imagens microscópicas preparadas com a técnica Kato-Katz

Abstract

A major public health concern is caused by human intestinal parasites, which are found largely in tropical countries. The diagnosis of these parasitic diseases is made through physiological symptoms and fecal examination. Often, few professionals are available and able to perform this type of examination, which is considered slow, difficult, error-prone, and can cause eye fatigue in the specialist. Artificial intelligence techniques have been successfully applied to problems of this nature. Therefore, the objective of this work is to develop a solution based on deep learning and machine learning to find intestinal parasite eggs of the species S. mansoni, being a system to aid decision-making in the diagnosis of fecal examination whose slides were prepared using the Kato-Katz parasitological technique. A real database was built with 1100 images that were annotated by three different human specialists in the diagnosis of schistosomiasis. Data augmentation techniques online and offline were used to obtain a larger number of samples and improve the generalizability of the tool. As a result, the proposed solution achieved an AP value of 0.884 for an @[IoU=0.50]. The results and employability of the system are promising, and it could be used in the SUS to assist health professionals in diagnosing schistosomiasis.